[gmx-users] berendsen P coupling and fluctuation properties
Justin A. Lemkul
jalemkul at vt.edu
Tue Aug 16 19:17:43 CEST 2011
Elisabeth wrote:
> Hello,
>
> Thank you. I am looking at potential energies to calculate vaporization
> heat. I wanted to know how the fact that berendsen does not lead to
> correct ensemble is affecting total potential energy of the system. I
> have an unclear image of "correct ensemble". Does this mean whatever
> output I am getting from my runs are unreliable?
>
The Berendsen coupling algorithms produce very narrow distributions of
temperature and pressure. These do not correspond to the correct distribution
of a true statistical mechanical ensemble; plot a histogram and you'll find that
the results are shockingly different between Berendsen and, say, Nose-Hoover for
T. Therefore, the ensemble you're using to measure your properties is not NPT
(or NVT, in the case of thermostat only), it's something undefined and likely
not real. Applying better thermostats and barostats that have been shown to
produce the correct distributions is more rigorously correct.
The effects are usually not terribly noticeable unless you go looking for them.
You'll find that the Berendsen algorithms quite faithfully give you the target
T and P that you desire (which does make them very good for initial
equilibration) and so you naturally assume that everything is fine. While
that's all well and good in one sense, the fluctuations of T (and thus
velocities, and thus kinetic energy) are wrong. If KE does not fluctuate
properly, then neither does PE (since you're then affecting energy flow between
PE and KE, while total energy stays fixed, in theory).
The larger point is that if you claim to apply an NVT (or NPT) ensemble using
Berendsen coupling, you're not correct, strictly speaking.
-Justin
> Appreciate any clarification.
>
> Thank you,
> Best,
>
> On 15 August 2011 19:34, Mark Abraham <Mark.Abraham at anu.edu.au
> <mailto:Mark.Abraham at anu.edu.au>> wrote:
>
> On 16/08/2011 7:05 AM, Elisabeth wrote:
>> Dear all,
>>
>> I noticed that applying Parrinello-Rahman (PR) pressure coupling
>> even after equilibration with berendsen does not lead to target
>> value for pressure when
>>
>> ; Bonds
>> constraints = none
>>
>> is used.
>
> The use of constraints and the integration step size is linked.
> Roughly speaking, no constraints should accompany a 0.5 fs time
> step, H-bond constraints with 1fs and all constraints with 2fs.
> Haphazard changes to .mdp files have all kinds of these "gotchas".
>
>
>> I tried Berendsen for to get fixed pressure ( 50 bar) but in the
>> next run PR even for long time is giving 52 bar. This is the case
>> for other target pressures too.
>
> You need to be sure to collect statistics only after equilibration,
> and consider whether the observed variation is consistent with
> convergence to a given value.
>
>
>>
>> So this made me select Berenden which is giving target pressure
>> values but my concern is whether my results are reliable because
>> BR does not give the exact ensemble as PR. I read somewhere on the
>> list that fluctuation properties can not be calculated when BR is
>> used. What does "fluctuation property" mean?
>
> BR does not produce the correct ensemble. I forget the details about
> why, but there are references in the T-coupling section of the
> manual you should consider.
>
>
>> Does this mean that any property calculated form fluctuations of
>> some other quantity can not be obtained>? like heat capacity which
>> is defined based on enthalpy fluctuations?
>
> IIRC, yes.
>
>
>> I am interested in potential energy terms (g_energy bonded/non
>> boned terms)
>
> I routinely struggle to see why people think they can learn anything
> from these.
>
>
>> and structural properties like rdf for a number of polymer
>> molecules, system size around 3000 atoms.
>>
>> Thank you for your comments.
>> Best,
>>
>> constraints = none
>>
>> ; Run control
>> integrator = md
>> dt = 0.001
>> nsteps = 5000000
>> nstcomm = 100
>>
>>
>> nstenergy = 100
>> nstxout = 100
>> nstlist = 10
>> ns_type = grid
>>
>> coulombtype = Shift
>> vdw-type = Shift
>> rcoulomb-switch = 0
>> rvdw-switch = 0.9 ;0
>>
>> ; Cut-offs
>> rlist = 1.25
>> rcoulomb = 1.0
>> rvdw = 1.0
>>
>> Tcoupl = v-rescale
>> tc-grps = System
>> tau_t = 0.1
>> ref_t = 300
>>
>> Pcoupl = berendsen
>> Pcoupltype = isotropic
>> tau_p = 1
>> compressibility = 3.5e-5
>> ref_p = 10
>>
>>
>> gen_vel = no
>> gen_temp = 300.0
>> gen_seed = 173529
>>
>>
>>
>>
>>
>>
>>
>
>
> --
> gmx-users mailing list gmx-users at gromacs.org
> <mailto:gmx-users at gromacs.org>
> http://lists.gromacs.org/mailman/listinfo/gmx-users
> Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/Search before posting!
> Please don't post (un)subscribe requests to the list. Use the
> www interface or send it to gmx-users-request at gromacs.org
> <mailto:gmx-users-request at gromacs.org>.
> Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>
>
--
========================================
Justin A. Lemkul
Ph.D. Candidate
ICTAS Doctoral Scholar
MILES-IGERT Trainee
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540) 231-9080
http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin
========================================
More information about the gromacs.org_gmx-users
mailing list